'''
classList = labels.tolist()
classDic = dict([(c, n) for n, c in enumerate(classList)])
classDic # In[]
#y_test= np.array([classDic[x] for x in y_testREAL])
y_pred = np.array([classDic[x] for x in yL])
cm_testREAL1 = confusion_matrix(y_testREAL1, y_pred)
#set(y_test)
audioUtils.plot_confusion_matrix(cm_testREAL1, classList) #, normalize=True)
# In[]
yPredProb = m.predict(X_testREAL1)
y_pred = yPredProb.argmax(axis=1)
cm_testREAL1 = confusion_matrix(y_testREAL1, y_pred)
#set(y_test)
audioUtils.plot_confusion_matrix(
cm_testREAL1, classList,
title='Confusion matrix: y_testREAL1') #, normalize=True)
# In[]
rec = recall_score(y_test, y_pred, average='micro')
#F1 - high if both recall and precision are high.
from sklearn.metrics import f1_score
f1 = f1_score(y_test, y_pred, average='micro')
print("Precision:\t", round(prec, 3))
print("Recall:\t\t", round(rec, 3))
print("F1:\t\t", round(f1, 3))
# Confusion Matrix
from sklearn.metrics import confusion_matrix
cm = confusion_matrix(y_test, y_pred)
import audioUtils
audioUtils.plot_confusion_matrix(cm,
categories,
normalize=False,
save_path=save_dir)
# SAVE MODEL SUMMARY and METRICS TO FILE
f = open(save_dir + "Metrics.txt", "a")
stringlist = []
model.summary(print_fn=lambda x: stringlist.append(x))
short_model_summary = "\n".join(stringlist)
f.write(short_model_summary + "\n\n")
f.write('Train loss:\t' + str(round(train_loss, 3)) + "\n")
f.write('Train accuracy:\t' + str((round(1 - train_error, 3))) + "\n")
f.write('Val loss:\t' + str(round(val_loss, 3)) + "\n")
f.write('Val accuracy:\t' + str((round(1 - val_error, 3))) + "\n")
f.write('Test loss:\t' + str(round(test_loss, 3)) + "\n")
'bed', 'bird', 'cat', 'dog', 'happy', 'house',
'marvin', 'sheila', 'tree', 'wow']
# In[ ]:
#20cmd
classes=['unknown', 'nine', 'yes', 'no', 'up', 'down', 'left', 'right', 'on', 'off', 'stop', 'go',
'zero', 'one', 'two', 'three', 'four', 'five', 'six', 'seven', 'eight']
# In[ ]:
audioUtils.plot_confusion_matrix(cmP,classes, normalize=True)
audioUtils.plot_confusion_matrix(cmFN,classes, normalize=True)
# In[ ]:
# In[ ]:
def get_confusion_matrix(y_test, y_pred, keywords):
cm = confusion_matrix(y_test, np.argmax(y_pred, 1))
keywordList = list(keywords.keys())
audioUtils.plot_confusion_matrix(cm, keywordList, normalize=True)
#35word, v1
#classes=['nine', 'yes', 'no', 'up', 'down', 'left', 'right', 'on', 'off', 'stop', 'go',
# 'zero', 'one', 'two', 'three', 'four', 'five', 'six', 'seven', 'eight',
# 'bed', 'bird', 'cat', 'dog', 'happy', 'house',
# 'marvin', 'sheila', 'tree', 'wow']
# In[38]:
#20cmd
#classes=['unknown', 'nine', 'yes', 'no', 'up', 'down', 'left', 'right', 'on', 'off', 'stop', 'go',
# 'zero', 'one', 'two', 'three', 'four', 'five', 'six', 'seven', 'eight']
# In[39]:
audioUtils.plot_confusion_matrix(cm, classes, normalize=False)
# In[39]:
# In[40]:
import pandas as pd
import kapre
[tf.__version__, kapre.__version__, pd.__version__, librosa.__version__]
# # Save model to .pb
# In[48]:
#model = SpeechModels.AttRNNSpeechModel(nCategs, samplingrate = sr, inputLength = None)#, rnn_func=L.LSTM)
#model.load_weights('model-attRNN.h5')